Log odds ratio - logistic regression

[Rochelle Zhang]

Dear all,

I was confused by the two webpages showing logistic regression output on the Log odds ratio

the first one is from UCLA stata site:

http://www.ats.ucla.edu/stat/stata/o...a_logistic.htm

it seems to me,
female | 1.482498 .they interpret 1.482 as the log odds.

( sorry, i could not copy and paste the output table in nice format)

the second one is also referenced in others' stata post : https://www3.nd.edu/~rwilliam/stats3...xedEffects.pdf here the author specifically requested log odds ratio by issuing
xtlogit, or

mother | 1.790388 .2857157 3.65 0.000 1.309513 2.447848

they take the bolded number above as log odds.

My basic question, is the coefficient output from running logit dep.var independent var is the log odds

or do we have to call
logit, or to get log odds

thanks,
Rochelle

[Roman Mostazir]

Dear all,

I was confused by the two webpages showing logistic regression output on the Log odds ratio

the first one is from UCLA stata site:

http://www.ats.ucla.edu/stat/stata/o...a_logistic.htm

it seems to me,
female | 1.482498 .they interpret 1.482 as the log odds.

This is the logarithm base 'e' of the odds. Calling 'or' after 'logit' will provide the odds ratio.

the second one is also referenced in others' stata post : https://www3.nd.edu/~rwilliam/stats3...xedEffects.pdf here the author specifically requested log odds ratio by issuing
xtlogit, or

mother | 1.790388 .2857157 3.65 0.000 1.309513 2.447848

they take the bolded number above as log odds.

Nope, it is in odds ratio and clearly written in the section beneath the Table. Also 'or' is called upon for producing odds ratio.

My basic question, is the coefficient output from running logit dep.var independent var is the log odds

or do we have to call
logit, or to get log odds

 logit dep.var independent var, or // Without 'or' you get only log odds. 'or' gives odds ratios

logistic dep.var independent var // With 'logistic' you don't need to call 'or', it reports the odds ratios

[Rochelle Zhang]

Thank you Roman for explaining it so clearly !!!

[Will Hauser]

If you ever get confused a fairly safe practice (but not fool proof) is to look for negative coefficients. Since odds ratios cannot be negative the presence of negative values among the coefficients is a sign that you're probably seeing log odds reported. Of course this assumes that at least one of the covariates has a negative relationship with the outcome which is likely but not a guarantee.